1 // SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
2 /* Copyright (C) 2019 Netronome Systems, Inc. */
4 #include <linux/if_arp.h>
5 #include <linux/init.h>
6 #include <linux/kernel.h>
7 #include <linux/module.h>
8 #include <linux/mpls.h>
9 #include <linux/rtnetlink.h>
10 #include <linux/skbuff.h>
11 #include <linux/tc_act/tc_mpls.h>
13 #include <net/netlink.h>
14 #include <net/pkt_sched.h>
15 #include <net/pkt_cls.h>
16 #include <net/tc_act/tc_mpls.h>
18 static struct tc_action_ops act_mpls_ops;
20 #define ACT_MPLS_TTL_DEFAULT 255
22 static __be32 tcf_mpls_get_lse(struct mpls_shim_hdr *lse,
23 struct tcf_mpls_params *p, bool set_bos)
28 new_lse = be32_to_cpu(lse->label_stack_entry);
30 if (p->tcfm_label != ACT_MPLS_LABEL_NOT_SET) {
31 new_lse &= ~MPLS_LS_LABEL_MASK;
32 new_lse |= p->tcfm_label << MPLS_LS_LABEL_SHIFT;
35 new_lse &= ~MPLS_LS_TTL_MASK;
36 new_lse |= p->tcfm_ttl << MPLS_LS_TTL_SHIFT;
38 if (p->tcfm_tc != ACT_MPLS_TC_NOT_SET) {
39 new_lse &= ~MPLS_LS_TC_MASK;
40 new_lse |= p->tcfm_tc << MPLS_LS_TC_SHIFT;
42 if (p->tcfm_bos != ACT_MPLS_BOS_NOT_SET) {
43 new_lse &= ~MPLS_LS_S_MASK;
44 new_lse |= p->tcfm_bos << MPLS_LS_S_SHIFT;
46 new_lse |= 1 << MPLS_LS_S_SHIFT;
49 return cpu_to_be32(new_lse);
52 static int tcf_mpls_act(struct sk_buff *skb, const struct tc_action *a,
53 struct tcf_result *res)
55 struct tcf_mpls *m = to_mpls(a);
56 struct tcf_mpls_params *p;
60 tcf_lastuse_update(&m->tcf_tm);
61 bstats_update(this_cpu_ptr(m->common.cpu_bstats), skb);
63 /* Ensure 'data' points at mac_header prior calling mpls manipulating
66 if (skb_at_tc_ingress(skb)) {
67 skb_push_rcsum(skb, skb->mac_len);
68 mac_len = skb->mac_len;
70 mac_len = skb_network_header(skb) - skb_mac_header(skb);
73 ret = READ_ONCE(m->tcf_action);
75 p = rcu_dereference_bh(m->mpls_p);
77 switch (p->tcfm_action) {
78 case TCA_MPLS_ACT_POP:
79 if (skb_mpls_pop(skb, p->tcfm_proto, mac_len,
80 skb->dev && skb->dev->type == ARPHRD_ETHER))
83 case TCA_MPLS_ACT_PUSH:
84 new_lse = tcf_mpls_get_lse(NULL, p, !eth_p_mpls(skb_protocol(skb, true)));
85 if (skb_mpls_push(skb, new_lse, p->tcfm_proto, mac_len,
86 skb->dev && skb->dev->type == ARPHRD_ETHER))
89 case TCA_MPLS_ACT_MAC_PUSH:
90 if (skb_vlan_tag_present(skb)) {
91 if (__vlan_insert_inner_tag(skb, skb->vlan_proto,
92 skb_vlan_tag_get(skb),
96 skb->protocol = skb->vlan_proto;
97 __vlan_hwaccel_clear_tag(skb);
100 new_lse = tcf_mpls_get_lse(NULL, p, mac_len ||
101 !eth_p_mpls(skb->protocol));
103 if (skb_mpls_push(skb, new_lse, p->tcfm_proto, 0, false))
106 case TCA_MPLS_ACT_MODIFY:
107 if (!pskb_may_pull(skb,
108 skb_network_offset(skb) + MPLS_HLEN))
110 new_lse = tcf_mpls_get_lse(mpls_hdr(skb), p, false);
111 if (skb_mpls_update_lse(skb, new_lse))
114 case TCA_MPLS_ACT_DEC_TTL:
115 if (skb_mpls_dec_ttl(skb))
120 if (skb_at_tc_ingress(skb))
121 skb_pull_rcsum(skb, skb->mac_len);
126 qstats_drop_inc(this_cpu_ptr(m->common.cpu_qstats));
130 static int valid_label(const struct nlattr *attr,
131 struct netlink_ext_ack *extack)
133 const u32 *label = nla_data(attr);
135 if (nla_len(attr) != sizeof(*label)) {
136 NL_SET_ERR_MSG_MOD(extack, "Invalid MPLS label length");
140 if (*label & ~MPLS_LABEL_MASK || *label == MPLS_LABEL_IMPLNULL) {
141 NL_SET_ERR_MSG_MOD(extack, "MPLS label out of range");
148 static const struct nla_policy mpls_policy[TCA_MPLS_MAX + 1] = {
149 [TCA_MPLS_PARMS] = NLA_POLICY_EXACT_LEN(sizeof(struct tc_mpls)),
150 [TCA_MPLS_PROTO] = { .type = NLA_U16 },
151 [TCA_MPLS_LABEL] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
153 [TCA_MPLS_TC] = NLA_POLICY_RANGE(NLA_U8, 0, 7),
154 [TCA_MPLS_TTL] = NLA_POLICY_MIN(NLA_U8, 1),
155 [TCA_MPLS_BOS] = NLA_POLICY_RANGE(NLA_U8, 0, 1),
158 static int tcf_mpls_init(struct net *net, struct nlattr *nla,
159 struct nlattr *est, struct tc_action **a,
160 struct tcf_proto *tp, u32 flags,
161 struct netlink_ext_ack *extack)
163 struct tc_action_net *tn = net_generic(net, act_mpls_ops.net_id);
164 bool bind = flags & TCA_ACT_FLAGS_BIND;
165 struct nlattr *tb[TCA_MPLS_MAX + 1];
166 struct tcf_chain *goto_ch = NULL;
167 struct tcf_mpls_params *p;
168 struct tc_mpls *parm;
176 NL_SET_ERR_MSG_MOD(extack, "Missing netlink attributes");
180 err = nla_parse_nested(tb, TCA_MPLS_MAX, nla, mpls_policy, extack);
184 if (!tb[TCA_MPLS_PARMS]) {
185 NL_SET_ERR_MSG_MOD(extack, "No MPLS params");
188 parm = nla_data(tb[TCA_MPLS_PARMS]);
191 err = tcf_idr_check_alloc(tn, &index, a, bind);
199 ret = tcf_idr_create(tn, index, est, a, &act_mpls_ops, bind,
202 tcf_idr_cleanup(tn, index);
207 } else if (!(flags & TCA_ACT_FLAGS_REPLACE)) {
208 tcf_idr_release(*a, bind);
212 /* Verify parameters against action type. */
213 switch (parm->m_action) {
214 case TCA_MPLS_ACT_POP:
215 if (!tb[TCA_MPLS_PROTO]) {
216 NL_SET_ERR_MSG_MOD(extack, "Protocol must be set for MPLS pop");
220 if (!eth_proto_is_802_3(nla_get_be16(tb[TCA_MPLS_PROTO]))) {
221 NL_SET_ERR_MSG_MOD(extack, "Invalid protocol type for MPLS pop");
225 if (tb[TCA_MPLS_LABEL] || tb[TCA_MPLS_TTL] || tb[TCA_MPLS_TC] ||
227 NL_SET_ERR_MSG_MOD(extack, "Label, TTL, TC or BOS cannot be used with MPLS pop");
232 case TCA_MPLS_ACT_DEC_TTL:
233 if (tb[TCA_MPLS_PROTO] || tb[TCA_MPLS_LABEL] ||
234 tb[TCA_MPLS_TTL] || tb[TCA_MPLS_TC] || tb[TCA_MPLS_BOS]) {
235 NL_SET_ERR_MSG_MOD(extack, "Label, TTL, TC, BOS or protocol cannot be used with MPLS dec_ttl");
240 case TCA_MPLS_ACT_PUSH:
241 case TCA_MPLS_ACT_MAC_PUSH:
242 if (!tb[TCA_MPLS_LABEL]) {
243 NL_SET_ERR_MSG_MOD(extack, "Label is required for MPLS push");
247 if (tb[TCA_MPLS_PROTO] &&
248 !eth_p_mpls(nla_get_be16(tb[TCA_MPLS_PROTO]))) {
249 NL_SET_ERR_MSG_MOD(extack, "Protocol must be an MPLS type for MPLS push");
250 err = -EPROTONOSUPPORT;
253 /* Push needs a TTL - if not specified, set a default value. */
254 if (!tb[TCA_MPLS_TTL]) {
255 #if IS_ENABLED(CONFIG_MPLS)
256 mpls_ttl = net->mpls.default_ttl ?
257 net->mpls.default_ttl : ACT_MPLS_TTL_DEFAULT;
259 mpls_ttl = ACT_MPLS_TTL_DEFAULT;
263 case TCA_MPLS_ACT_MODIFY:
264 if (tb[TCA_MPLS_PROTO]) {
265 NL_SET_ERR_MSG_MOD(extack, "Protocol cannot be used with MPLS modify");
271 NL_SET_ERR_MSG_MOD(extack, "Unknown MPLS action");
276 err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
282 p = kzalloc(sizeof(*p), GFP_KERNEL);
288 p->tcfm_action = parm->m_action;
289 p->tcfm_label = tb[TCA_MPLS_LABEL] ? nla_get_u32(tb[TCA_MPLS_LABEL]) :
290 ACT_MPLS_LABEL_NOT_SET;
291 p->tcfm_tc = tb[TCA_MPLS_TC] ? nla_get_u8(tb[TCA_MPLS_TC]) :
293 p->tcfm_ttl = tb[TCA_MPLS_TTL] ? nla_get_u8(tb[TCA_MPLS_TTL]) :
295 p->tcfm_bos = tb[TCA_MPLS_BOS] ? nla_get_u8(tb[TCA_MPLS_BOS]) :
296 ACT_MPLS_BOS_NOT_SET;
297 p->tcfm_proto = tb[TCA_MPLS_PROTO] ? nla_get_be16(tb[TCA_MPLS_PROTO]) :
298 htons(ETH_P_MPLS_UC);
300 spin_lock_bh(&m->tcf_lock);
301 goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
302 p = rcu_replace_pointer(m->mpls_p, p, lockdep_is_held(&m->tcf_lock));
303 spin_unlock_bh(&m->tcf_lock);
306 tcf_chain_put_by_act(goto_ch);
313 tcf_chain_put_by_act(goto_ch);
315 tcf_idr_release(*a, bind);
319 static void tcf_mpls_cleanup(struct tc_action *a)
321 struct tcf_mpls *m = to_mpls(a);
322 struct tcf_mpls_params *p;
324 p = rcu_dereference_protected(m->mpls_p, 1);
329 static int tcf_mpls_dump(struct sk_buff *skb, struct tc_action *a,
332 unsigned char *b = skb_tail_pointer(skb);
333 struct tcf_mpls *m = to_mpls(a);
334 struct tcf_mpls_params *p;
335 struct tc_mpls opt = {
336 .index = m->tcf_index,
337 .refcnt = refcount_read(&m->tcf_refcnt) - ref,
338 .bindcnt = atomic_read(&m->tcf_bindcnt) - bind,
342 spin_lock_bh(&m->tcf_lock);
343 opt.action = m->tcf_action;
344 p = rcu_dereference_protected(m->mpls_p, lockdep_is_held(&m->tcf_lock));
345 opt.m_action = p->tcfm_action;
347 if (nla_put(skb, TCA_MPLS_PARMS, sizeof(opt), &opt))
348 goto nla_put_failure;
350 if (p->tcfm_label != ACT_MPLS_LABEL_NOT_SET &&
351 nla_put_u32(skb, TCA_MPLS_LABEL, p->tcfm_label))
352 goto nla_put_failure;
354 if (p->tcfm_tc != ACT_MPLS_TC_NOT_SET &&
355 nla_put_u8(skb, TCA_MPLS_TC, p->tcfm_tc))
356 goto nla_put_failure;
358 if (p->tcfm_ttl && nla_put_u8(skb, TCA_MPLS_TTL, p->tcfm_ttl))
359 goto nla_put_failure;
361 if (p->tcfm_bos != ACT_MPLS_BOS_NOT_SET &&
362 nla_put_u8(skb, TCA_MPLS_BOS, p->tcfm_bos))
363 goto nla_put_failure;
365 if (nla_put_be16(skb, TCA_MPLS_PROTO, p->tcfm_proto))
366 goto nla_put_failure;
368 tcf_tm_dump(&t, &m->tcf_tm);
370 if (nla_put_64bit(skb, TCA_MPLS_TM, sizeof(t), &t, TCA_MPLS_PAD))
371 goto nla_put_failure;
373 spin_unlock_bh(&m->tcf_lock);
378 spin_unlock_bh(&m->tcf_lock);
383 static int tcf_mpls_offload_act_setup(struct tc_action *act, void *entry_data,
384 u32 *index_inc, bool bind,
385 struct netlink_ext_ack *extack)
388 struct flow_action_entry *entry = entry_data;
390 switch (tcf_mpls_action(act)) {
391 case TCA_MPLS_ACT_PUSH:
392 entry->id = FLOW_ACTION_MPLS_PUSH;
393 entry->mpls_push.proto = tcf_mpls_proto(act);
394 entry->mpls_push.label = tcf_mpls_label(act);
395 entry->mpls_push.tc = tcf_mpls_tc(act);
396 entry->mpls_push.bos = tcf_mpls_bos(act);
397 entry->mpls_push.ttl = tcf_mpls_ttl(act);
399 case TCA_MPLS_ACT_POP:
400 entry->id = FLOW_ACTION_MPLS_POP;
401 entry->mpls_pop.proto = tcf_mpls_proto(act);
403 case TCA_MPLS_ACT_MODIFY:
404 entry->id = FLOW_ACTION_MPLS_MANGLE;
405 entry->mpls_mangle.label = tcf_mpls_label(act);
406 entry->mpls_mangle.tc = tcf_mpls_tc(act);
407 entry->mpls_mangle.bos = tcf_mpls_bos(act);
408 entry->mpls_mangle.ttl = tcf_mpls_ttl(act);
410 case TCA_MPLS_ACT_DEC_TTL:
411 NL_SET_ERR_MSG_MOD(extack, "Offload not supported when \"dec_ttl\" option is used");
413 case TCA_MPLS_ACT_MAC_PUSH:
414 NL_SET_ERR_MSG_MOD(extack, "Offload not supported when \"mac_push\" option is used");
417 NL_SET_ERR_MSG_MOD(extack, "Unsupported MPLS mode offload");
422 struct flow_offload_action *fl_action = entry_data;
424 switch (tcf_mpls_action(act)) {
425 case TCA_MPLS_ACT_PUSH:
426 fl_action->id = FLOW_ACTION_MPLS_PUSH;
428 case TCA_MPLS_ACT_POP:
429 fl_action->id = FLOW_ACTION_MPLS_POP;
431 case TCA_MPLS_ACT_MODIFY:
432 fl_action->id = FLOW_ACTION_MPLS_MANGLE;
442 static struct tc_action_ops act_mpls_ops = {
445 .owner = THIS_MODULE,
447 .dump = tcf_mpls_dump,
448 .init = tcf_mpls_init,
449 .cleanup = tcf_mpls_cleanup,
450 .offload_act_setup = tcf_mpls_offload_act_setup,
451 .size = sizeof(struct tcf_mpls),
454 static __net_init int mpls_init_net(struct net *net)
456 struct tc_action_net *tn = net_generic(net, act_mpls_ops.net_id);
458 return tc_action_net_init(net, tn, &act_mpls_ops);
461 static void __net_exit mpls_exit_net(struct list_head *net_list)
463 tc_action_net_exit(net_list, act_mpls_ops.net_id);
466 static struct pernet_operations mpls_net_ops = {
467 .init = mpls_init_net,
468 .exit_batch = mpls_exit_net,
469 .id = &act_mpls_ops.net_id,
470 .size = sizeof(struct tc_action_net),
473 static int __init mpls_init_module(void)
475 return tcf_register_action(&act_mpls_ops, &mpls_net_ops);
478 static void __exit mpls_cleanup_module(void)
480 tcf_unregister_action(&act_mpls_ops, &mpls_net_ops);
483 module_init(mpls_init_module);
484 module_exit(mpls_cleanup_module);
486 MODULE_SOFTDEP("post: mpls_gso");
487 MODULE_AUTHOR("Netronome Systems <oss-drivers@netronome.com>");
488 MODULE_LICENSE("GPL");
489 MODULE_DESCRIPTION("MPLS manipulation actions");